Objectivity is a Unicorn

citations
gm
local
maize
policy
Author

Dan Hicks

Published

September 6, 2013

Following up on my last post and a semi-related conversation with a new officemate, it seems to me that a lot of people might react something like this:

Well, no surprise that we can’t trust these scientists: They have ties to Monsanto. What we need are objective, disinterested scientists who aren’t dependent on industry sponsors.

In this post I’m going to criticize this response. As the title puts it, objectivity is a unicorn. It doesn’t exist, it’s a myth, and so it’s not going to help us solve the problem.

That’s hyperbolic, for a few reasons. First, there are a few things we might mean by “objectivity.” Over the past few decades, philosophers have developed several different conceptions of objectivity. The one I’m targeting here is a “classical” notion of objectivity as disinterestedness or value-freedom. Specifically, I’m going to argue that the way we answer factual, scientific questions (like whether GM maize is more productive) depend on value judgments.1 Second, there are a few areas of science where this isn’t so much the case and it’s (more or less) possible to achieve disinterestedness. Pure mathematics and theoretical physics are good examples — though even here we might have questions about “physics fundamentalism.”

To make things concrete, I’m going to focus on one seemingly-simple, factual question: Is GM maize more productive than non-GM maize? My argument, in brief, is that, whenever we investigate something as complex as the food system, before we answer this question we need to settle numerous methodological issues — we need to make a lot of assumptions. And the way we settle these issues — the specific assumptions scientists make when they try to answer this question — depend on value judgments. Hence, the way we answer this question depends on value judgments.2

Indeed, when it comes to questions like the productivity of GM maize, the value judgments are often deeply controversial. As I’ll argue below, the assumptions scientists make when they evaluate crop productivity are closely related to assumptions about the kind of food system we should have. The upshot is that we need something like what philosopher Matt Brown calls “the democratic control of the scientific control of politics,” which brings scientific and political concerns together more deliberately and (hopefully) constructively, rather than — as the suggestion at the top of the post would have it — trying to separate them. Rather than chasing mythical objectivity, we need new institutions that integrate science and democracy.

Now to the details of the argument. Again, I’m looking at one seemingly-simple, factual question: Is GM maize more productive than non-GM maize? There are at least five different kinds of methodological issues and assumptions that we need to settle before we can answer this question.

The first kind of issue to settle is which GM traits. The two most commonly used GM traits are resistance to the herbicide glyphosate (RoundUp, manufactured by Monsanto) and the production of an insecticide using genes from a bacterium (this trait is called Bt, for the scientific name of the bacterium). (In the last post, the paper was looking at Bt maize and cotton.) Other traits that are available for farmers to grow today and in the near future, but aren’t used as much, include drought tolerance, disease resistance, and enhanced nutritional content. Clearly, the productivity of GM maize depends a lot on exactly which traits we’re talking about — drought tolerance might boost productivity a lot, but herbicide-tolerance probably won’t.

The second kind of issue is the range of ecological contexts for the maize. Different places have different insect populations, and the Bt toxin is effective against some kinds of pests but not others. Likewise with the presence of pathogens. Water, soil chemistry and structure, and temperature (both in general and over the course of a particular growing season) can make a big difference. And maize can produce more or less in the presence of various kinds of plants — the closer the maize seeds are planted the more plants per unit of land, but if they’re too close they can be smaller or less productive per plant; weeds might produce soil toxins or shade maize sprouts that need lots of sunlight. There’s also the question of whether we’re looking at monocultured maize — big fields of nearly-genetically identical plants — or polycultures — which typically involve smaller fields and growing several different kinds of plants together.

These ecological contexts can be managed by various agricultural practices, which is the third set of issues. Indeed, from a certain perspective, farming just is managing ecology. At this point, the dependence on value judgments starts to become obvious. If we’re very concerned about the environmental impacts of agriculture — things like fertilizer runoff, the effects of pesticides on non-targeted insects and other animals, and the use of fossil fuels to grow and transport food — we might value organic over conventional, industrialized agriculture. Or, if we’re employees of a company like Monsanto, we might have a strong values-based preference for the conventional, industrial status quo. Locavores and traditionalists value an agricultural system with many small-scale farmers using inexpensive but labor-intensive methods; while people who are most concerned about economic efficiency would value a system with fewer, large-scale farmers using advanced (and expensive) technology to minimize labor expenses. There are also relevant differences between the highly-industrialized, commercial system in core countries like Canada and the US; semi-industrialized commercial agriculture in periphery and semi-periphery regions; and subsistence agriculture in periphery and maginal regions. For example, access to capital — and so access to advanced technology — varies widely between (and within) these three different systems.

It might seem like, since GM maize is going to be grown in a highly-industrialized commercial system, these questions are irrelevant. But the productivity question is comparative: is GM maize more productive than the non-GM alternative? And so the way we answer this question depends on what the non-GM alternatives are. GM maize is going to be more productive than maize grown in subsistence agriculture, but because it’s grown in a more intensive agricultural system, not because it’s genetically modified. Once we raise the possibility of these kinds of comparisons — which I think we should — we should also raise the possibility that the productivity question isn’t even the right question to ask.

Returning to our list, a closely-related fourth set of issues concerns political economy: Who owns the farmland, who decides how it’s farmed, and what happens to the maize after it’s harvested? For instance, is this maize being grown as a staple food for humans, or will it be processed into high fructose corn syrup and livestock feed? Again, while these issues might not seem directly relevant to the productivity question, they raise issues of what alternatives we’re comparing the GM maize against and what kind of broader food system we’re assuming.

The question of use is also closely related to the fifth and perhaps most scientifically fundamental set of issues, measurement conventions. We can take any ratio between outputs and inputs to be our measure of productivity. And there are a lot of different ways we might measure the outputs of GM maize: tons (weight), bushels (volume), farmer income (economic), human-edible calories of the raw maize, human-edible calories after processing, cattle-edible calories (cattle can digest parts of the plant that we can’t), and so on. Likewise, there are a lot of different ways we might measure inputs: per acre (land), per weight of chemical inputs, per farmer expenses (capital), per hour of worker labor. Bushels per acre or hectare are common measures, but if we’re interested in feeding the world and the environmental impacts of food production it would probably make more sense to look at calories produced using various land and chemical inputs.

Once we’ve decided how we’re going to measure productivity, we need to decide how to handle variation and uncertainty. Productivity will not be exactly the same in all places and times. Will we report an average, or some more complicated piece of statistics, such as a confidence interval or regression analysis? The more complicated statistics can give us a more sophisticated understanding of how productive GM maize is, but themselves rely on assumptions that bring in further value-judgments. For example, if we decide to use a confidence interval, we need to decide what confidence levels to use; and this decision requires taking various risks into account.3 We also need to take into account the fact that no measurement is perfect. Statistical techniques can help to reduce uncertainty but not completely eliminate it, and these techniques often involve tradeoffs among different kinds of uncertainty. We need to decide how much of each kind of uncertainty we can tolerate, which depends on our purposes — the reasons why we’re asking whether GM maize is more productive in the first place — and so depends on the value-judgments that are motivating the research.

In this post, I focused on a single, seemingly-simple, factual question: Is GM maize more productive than non-GM maize? I showed how answering this question depends on settling at least five different kinds of issues or assumptions: the kinds of GM traits, the range of ecological contexts and agricultural practices, the political economy of food production, and measurement conventions. For each of these, I argued that the complexity of the issue requires bringing in value judgments before we can answer the productivity question. Thus an objective answer to the question — in the sense of disinterestedness or value-freedom — is impossible. Objectivity is as mythical as a unicorn. Consequently, we need to find alternative ways to manage the relationship between science and policy.


  1. For philosophers: I’m writing this post for a non-philosophical audience, so I’m not going to spell out “value judgments” in more detail, or distinguish epistemic, internal, or constitutive from their usual constrasts. As someone with one appendange in ethics, I actually think the way philosophers of science talk about “values” is problematic. ↩︎

  2. By no means is this argument original. One nice version that’s probably accessible for non-academics is here. ↩︎

  3. For more on risk, values, and the relationship between science and policy, check out the work of philosopher Heather Douglas. ↩︎